Skip to content
WattCurious

Why Is My Solar System Producing Less Power Than Expected?

Updated 8 October 20267 min readSolar Energy

A system that underperforms is usually doing one of two things: producing exactly what its conditions allow while the expectation was wrong, or losing output to something specific and findable. Telling those apart is the whole job, and it is done in that order — establish what the day should have given, then look at the shape of what it actually gave, then compare the system against itself.

Key takeaways

  • A panel's rating is measured under laboratory conditions that almost never occur outdoors, so real output starts below it before anything goes wrong.
  • Ordinary losses — heat, soiling, wiring, inverter conversion, angle of incidence — multiply rather than add, and together they are large.
  • The shape of the day curve identifies the problem faster than the daily total does.
  • A system with several strings carries its own control group: compare them against each other under the same sky.
  • Real faults announce themselves as a change over time or a difference between identical parts, not as a number that is merely lower than hoped.
On this page
  1. First decide what "expected" means
  2. The losses that are always there
  3. Read the shape, not the total
  4. The system is its own control group
  5. What actually reduces output
  6. Working through it in order

First decide what "expected" means

The number on a module is measured in a laboratory under standard test conditions: full-strength light arriving perpendicular to the cell, at a cell temperature of 25 °C (77 °F), with a defined atmospheric path. Those conditions are not a typical day. They are barely any day.

A cascade from a module's nameplate rating down to energy delivered at the meter. Each stage removes a share: light arriving at an angle rather than straight on, cell temperature above the rating condition, soiling on the glass, mismatch between modules and losses in the wiring, conversion in the inverter, and losses between the inverter and the meter. The bar at the end is substantially shorter than the one at the start, and the caption notes that none of these stages is a fault.A cascade from a module's nameplate rating down to energy delivered at the meter. Each stage removes a share: light arriving at an angle rather than straight on, cell temperature above the rating condition, soiling on the glass, mismatch between modules and losses in the wiring, conversion in the inverter, and losses between the inverter and the meter. The bar at the end is substantially shorter than the one at the start, and the caption notes that none of these stages is a fault.
Every stage takes a percentage and they multiply rather than add. Nothing in this cascade is broken — this is what a healthy system looks like.

So a shortfall against the nameplate is the normal state, and the useful question is never "is it below the rating" but "is it below what this site, this orientation and this weather should give". That expectation comes from a model of the site — the same irradiance data sets that make off-grid load sizing possible — or, more practically, from the system's own history on comparable days.

A great many reported faults dissolve at this step. The system was never going to produce the number it was measured against.

The losses that are always there

Five effects account for most of the gap, and all five are present on a system in perfect condition.

Temperature. Cells lose voltage as they heat, and a module in the sun runs well above air temperature — the mechanism behind efficiency loss in hot weather. On a hot afternoon this is the largest single loss in the chain.

Angle of incidence. Light striking the glass obliquely is partly reflected before it ever enters, on top of the cosine loss that governs the best tilt and direction for an array. Morning and evening pay this heavily; midday barely at all.

Soiling. A slow, site-specific accumulation that recovers after rain, described in cleaning panels without damaging them.

Mismatch and wiring. Modules in a string are not identical, and the string is limited by its weakest member; cable resistance takes a further slice on the way to the inverter.

Conversion. The inverter is efficient but not perfect, and least efficient at very low input — which is why the first and last hour of the day convert worse than the middle.

Read the shape, not the total

A daily total tells you that something is wrong. The shape of the day tells you what.

Five output curves across one day, compared against a smooth clear-sky reference. A normal clear day rises and falls symmetrically. A clipped day has a flat top where the inverter holds its rated output. A shaded day has a sharp notch at the same time each day with the rest of the curve intact. A cloudy day is ragged and irregular with sharp peaks and troughs. A failed-string day has the normal shape at reduced height across the whole day.Five output curves across one day, compared against a smooth clear-sky reference. A normal clear day rises and falls symmetrically. A clipped day has a flat top where the inverter holds its rated output. A shaded day has a sharp notch at the same time each day with the rest of the curve intact. A cloudy day is ragged and irregular with sharp peaks and troughs. A failed-string day has the normal shape at reduced height across the whole day.
Five different problems, five different shapes. The notch, the flat top, the ragged trace and the uniformly lower curve each point somewhere specific.

A flat top is clipping: the array is briefly offering more than the inverter converts, and the inverter holds at its limit. Normal, and deliberate. A notch at the same clock time every day is shade, because shadows keep a schedule — the subject of how partial shade affects panels. A ragged trace is cloud, and on some days bright cloud edges briefly push output above the clear-sky line. A curve of normal shape but uniformly lower points at something affecting the whole array equally: a string offline, a derating inverter, or soiling. And a curve that stops early or starts late is an obstruction on the horizon rather than anything electrical.

The system is its own control group

Any array with more than one string carries a built-in comparison, and it is the most informative measurement available without instruments or reference data.

Three strings on one inverter, reported side by side at the same moment. Two strings show closely matching voltage and current. The third shows the same voltage but noticeably lower current, which points to a shaded, soiled or partially disconnected string rather than to weather. A second panel shows the opposite case: one string with normal current but low voltage, pointing at a module or connection within that string.Three strings on one inverter, reported side by side at the same moment. Two strings show closely matching voltage and current. The third shows the same voltage but noticeably lower current, which points to a shaded, soiled or partially disconnected string rather than to weather. A second panel shows the opposite case: one string with normal current but low voltage, pointing at a module or connection within that string.
Same sky, same moment, same hardware. A difference between strings cannot be blamed on the weather, which is exactly what makes the comparison worth more than any absolute number.

Read the strings at the same moment, under the same sky. Matching strings mean the array is behaving consistently and any shortfall is a whole-system or expectation problem. One string low on current while its voltage matches points at shading, soiling or a fault in that string's path. One string low on voltage while current matches points at a module or a connection within it. And a string reading zero is a disconnection, a blown string fuse or a tripped isolator — a component, not a condition.

What actually reduces output

Once the expectation is right and the shape is understood, the list of real causes is short.

Symptom, likely cause, and what to check
What you seeMost likelyCheck
Uniformly lower, all day, every dayExpectation, not faultModel the site; compare with the system's own history
Gradual decline over weeksSoilingRainfall since the last clean; the size of the recovery after rain
Notch at the same time dailyShadeWhat casts it at that hour, in that season
Sudden permanent step downA string or input offlinePer-string current; fuses and isolators
Flat top around middayClippingWhether the flat section is wider than designed
Worse only when hotTemperature, or ventilationClearance behind modules; roof-surface mounting
One string low, others normalThat stringShade, soiling, a connector, a module
Inverter reporting reduced outputDeratingVentilation and ambient temperature at the inverter
Slow decline across yearsDegradationCompare like seasons, years apart

A sequence for narrowing the search, not a repair procedure. Anything behind an enclosure, or on the AC side, belongs to a qualified installer — the array is a live DC source whenever there is daylight on it.

Two entries deserve emphasis. Derating is an inverter protecting itself: too hot, or asked for more than it can deliver, it reduces output deliberately rather than failing. An inverter in direct sun, in a hot roof space, or with its vents blocked will do this every summer afternoon and report it as normal operation, because it is.

Three traces of output over time, each normalised against the weather. The first drifts gently downward over several years with seasonal ripple, labelled degradation. The second sawtooths downward over weeks and recovers sharply after rain, labelled soiling. The third holds steady and then drops vertically to a lower level and stays there, labelled a component going offline. A note says the time axis tells them apart.Three traces of output over time, each normalised against the weather. The first drifts gently downward over several years with seasonal ripple, labelled degradation. The second sawtooths downward over weeks and recovers sharply after rain, labelled soiling. The third holds steady and then drops vertically to a lower level and stays there, labelled a component going offline. A note says the time axis tells them apart.
Years, weeks, or an instant. The rate of a decline identifies its cause before any hardware is opened.

And connector resistance is the quiet one. A corroded or poorly seated DC connector adds resistance, which costs a little output continuously and generates heat at the joint. It rarely shows as a dramatic drop and often shows first as one string slightly behind its neighbours.

Working through it in order

The sequence matters more than any individual test, because each step removes a class of explanation.

A five-step sequence. First, establish the expected output for the site and the weather. Second, look at the shape of the day curve rather than the daily total. Third, compare strings against each other at the same moment. Fourth, check what changed and when it changed. Fifth, inspect the specific candidates the previous steps point at. Each step is labelled with the class of explanation it eliminates.A five-step sequence. First, establish the expected output for the site and the weather. Second, look at the shape of the day curve rather than the daily total. Third, compare strings against each other at the same moment. Fourth, check what changed and when it changed. Fifth, inspect the specific candidates the previous steps point at. Each step is labelled with the class of explanation it eliminates.
Each step removes a whole class of explanation. Starting at the last one — inspecting hardware — is how afternoons disappear.

Establish the expectation. Look at the shape. Compare the strings. Ask what changed, and when — a step on the day of a storm points straight at hail damage, visible or not. Only then inspect hardware, and only the candidates the first four steps left standing.

The reason this order works is that the common causes are cheap to rule out and the rare ones are expensive to investigate. A system that turns out to be producing exactly what a hot, hazy, slightly dusty week allows is the most common outcome of all — and the fastest one to reach, provided nobody started on the roof.

Frequently asked questions

How much less than the panel rating should I expect?

Less, always, and by a margin that is a property of the site rather than the equipment. The rating is measured at a cell temperature of 25 °C (77 °F) with light arriving straight on at full strength — conditions a working roof rarely meets. Every stage between the cell and the meter then takes a percentage. The right expectation comes from a model of your site, not from the sticker.

My output dropped suddenly. Where do I start?

With when it dropped and whether it affects everything equally. A sudden, permanent step points at a component: a string offline, a blown fuse, an inverter derating or shutting down part of its input. A drop that appears at the same time each day points at shade. A gradual drift over weeks points at soiling. The timing narrows it faster than any measurement.

Is a flat top on the output curve a fault?

Usually not. It is clipping: the array is briefly offering more power than the inverter is rated to convert, so the inverter holds its output at its limit. That is a design decision — an inverter matched to the rare peak would spend its life underused — and the energy given up is small. It becomes worth investigating only if the flat section is much wider than the design intended.

Can panels just wear out?

They degrade, slowly and predictably, losing a small fraction of their output per year through mechanisms in the cell and the encapsulant. Over a year that is invisible against weather; over a decade it is real. Degradation is a gentle downward slope across years, so it never explains a change that happened last week.

Do I need monitoring equipment to diagnose this?

It makes everything easier, but the essentials are readable without it. Most inverters report per-string voltage and current, and comparing strings against each other at the same moment is the single most informative measurement available — no reference data and no instruments required.

Sources

Named organisations whose published material underpins this article. Where no link is given, the source is named rather than linked.

Editorial Team

Research, drafting and review

Articles are drafted from primary engineering and physics references with AI-assisted tools, then reviewed and fact-checked line by line by a human editor before publication. We publish explanations, not recommendations: no products, no pricing, no country-specific rules, and no invented author personas.

Last reviewed 8 October 2026. How we research and review